Lately, we’ve talked a lot about the importance of ensuring that governments take into account open source, especially when it comes to AI. We submitted comments to NIST on Dual-Use Foundation Models and NTIA on the benefits of openness, and advocated in Congress. As frontier models and big tech continue to dominate the policy discussion, we need to ensure that open source remains top of mind for policymakers and regulators. At Mozilla, we know that open source is a fundamental driver of software that benefits people instead of a few big tech corporations, and it helps enable breakthroughs in medicine, science, and allows smaller companies to compete with tech giants. That’s why we’ll continue to raise the voice of the open source community in regulatory circles whenever we can – and most recently, at the Department of Commerce.
Last month, the Bureau of Industry and Security (BIS) released a proposed rule about reporting requirements for developing advanced AI models and computing clusters. This rule stems from the White House’s 2023 Executive Order on AI, which focuses on the safe and trustworthy development of AI. BIS asked for feedback from industry and stakeholders on topics such as the notification schedule for entities covered by the rule, how information is collected and stored, and what thresholds would trigger reporting requirements for these AI models and clusters.
While BIS’ proposed rule seeks to balance national security with economic concerns, it doesn’t adequately take into account the needs of the open source community or provide clarity as to how the proposed rule may affect them. This is critical given some of the most capable and widely used AI models are open source or partially open source. Open source software is a key driver of technological progress in AI and creates tremendous economic and security benefits for the United States. In our full comments, we set out how BIS can further engage with the open-source community and we emphasize the value that open-source offers for both the economy and national security. Below are some key points from our feedback to BIS:
1. BIS should clarify how the proposed rules would apply to open-source projects, especially since many don’t have a specific owner, are distributed globally, and are freely available. Ideally BIS could work with organizations like the Open Source Initiative (OSI) to come up with a framework.
2. As BIS updates the technical conditions for collection thresholds in response to technological advancements, we suggest setting a minimum update cycle of six months. This is crucial given the rapid pace of change in the AI landscape. It’s also necessary to maintain BIS’ core focus on the regulation of frontier models and to not unnecessarily stymie innovation across the broader AI ecosystem.
3. BIS should provide additional clarity about what ‘planned applicable activities’ and when a project is considered ‘planned.’
Mozilla appreciates BIS’s efforts to try and balance the benefits and risks of AI when it comes to national and economic security. We hope that BIS further considers the potential impact of the proposed rule and future regulatory actions on the open source community and appropriately weighs the myriad benefits which open source AI and open source software more broadly produce for America’s national and economic interests. We look forward to providing views as the US Government continues work on these important issues.
The post Mozilla Responds to BIS’ Proposed Rule on Reporting Requirements for the Development of Advanced AI Models and Computing Clusters appeared first on Open Policy & Advocacy.